Many fish species occur in areas with complicated geography. Natural barriers such as islands and coastlines mean that the spatial structure of the population is unlikely to be stationary. Here I develop and fit a spatiotemporal model that accounts for nonstationarity. The stochastic partial differential equation approach is used to reduce the computational burden. A simulation study demonstrates improved abundance estimates. This improvement has the potential to improve management decisions by more accurately reflecting a stock’s spatial structure. It should also provide more trustworthy estimates of uncertainty. These combined have the potential to improve management decision in many fisheries.